American Incarceration Is Mostly “Short” Terms, Including In Mississippi.

In my other article, I claimed:

The reality is that the vast majority of American prison terms — perhaps 80% of them, probably more — last less than five years before release or parole. Most prison terms last less than two years, and I doubt that even 5% last continuously for twenty years or more. I’ll support this more in another article, but it’s not a controversial claim, even though incarceration statistics can be messy and federal numbers might push it up a bit.

I don’t think anyone who’s studied incarceration would seriously dispute that. But just in case, and to try looking away from the usual data sources, I wrote a scraper that pulled every current inmate in Mississippi, an unusually punitive state, from their online database.

Limiting inmate locations to the ones on this list, except for the “restitution centers”, gave 18,603 inmates compared to the most recent daily inmate population of 18,695. I found or estimated admission years for 18,421 of them by using “Entry Date” or by subtracting “Total [Sentence] Length” from “Tentative Release Date”, so I’ve probably covered at least 98% of the Mississippi prison population.

The results: Half of these Mississippi inmates were admitted after 2012; 63.9% were admitted after 2010, and only 4.4% of them— 802 individuals — were admitted before 1996 (with 56% of those convicted of murder). Throw in every single inmate whose admission date I can’t guess (which does include a lot of life sentences) or which might be in a location I’m not checking and you might get it to 5.8%.

In other words: Imposing a maximal “20-year term”, right now, would reduce Mississippi’s “prison population” by all of 4% to 6% — a pretty marginal change.

To try to get a sense of the distribution of time served across actual people, I looked at Mississippi’s new prison terms (new court commitments) by year from the NPS and compared the number of people admitted in a given year with the number of remaining inmates who were admitted then. This will be an over-count, since the “entry date” of current Mississippi inmates might include people who returned to prison after parole violations. Also keep in mind that people enter prison throughout the year.

This is shown in the below chart, where black is current inmates by “entry date” year and orange is admissions by year minus current inmates, with non-”new court commitment” admissions outlined in grey.

With the above caveats in mind, at most 32.4% of 2013 Mississippi “new court commitments” have served continuously since then, and the numbers for 2011 and 1996 are at most 15.6% and 4% respectively. Barring some pretty dramatic sentencing changes, that means that probably around 68% of Mississippi inmates serve under three years, around 84% serve under five years, and at most around 4% serve for twenty years or more before release or parole.

Despite potential issues with admissions throughout the year and paroles, this is quite consistent with other states and other sources. A decade ago, Florida estimated that 69% of inmates admitted in FY2003–2004 would serve under three years, 83.5% would serve under five years, and just 3.2% would serve for twenty years or more. When Neal and Rick aggregated NCRP numbers across California, Colorado, Michigan, New Jersey, North Dakota, South Carolina, Washington, and Wisconsin, they found that 66% of prison terms that started in 2000 ended within 2 years, 87% ended within 5 years, and 93% ended within 10 years. Federal prison terms seem to be longer than state prison terms, but there are a lot fewer of them, and the median time served for federal prisoners released around 2012 was still just under 2 years.

Based on all that, I’m pretty confident that my earlier claim is, if anything, conservative: At least 80% of prison terms are under five years, most are under two years, and I doubt that even 5% run continuously for twenty years or more. It would be interesting to know more about how parole violations and recidivism worked in Mississippi, though.